The present invention relates to a system for high-frequency evaluation of probe measurement networks and, in particular, to a system for accurately evaluating the signal conditions existing in such networks even in those ones of such networks, for example, that are of a multichannel type in which each channel communicates through a separate device-probing end and even in those ones of such multichannel networks, for example, that have their device-probing ends crowded together in a high-density coplanar probing array as suitable for the measurement of integrated circuits or other microelectronic devices.
FIG. 1 shows a probe station 20 that includes a multichannel measurement network 21 of a type suitable for measuring high-frequency microelectronic devices at the wafer level. A probe station of this type is manufactured, for example, by Cascade Microtech, Inc. of Beaverton, Oreg. and sold under the trade name SUMMIT 10000. The various devices 24, the characteristics of which are to be measured by the network, are formed on the surface of a wafer 22 in isolation from each other. An enlarged schematic plan view of an individual device 24 is shown in FIG. 2. The surface of each device includes a predetermined pattern of bonding pads 26 that provide points of connection to the respective electrical components (not shown) formed on the central area of each device. The size of each bonding pad is exaggerated for ease of illustration in FIG. 2, but it will be recognized by one of ordinary skill in the art that there will typically be hundreds of bonding pads in the rectangular arrangement shown, each of a size that is barely visible to the eye without magnification. If a hybrid device instead of a flat wafer is being tested, then the individual devices can rise to different heights above the plane of the hybrid device's upper surface.
As depicted in FIG. 1, to facilitate high-frequency measurement of each device 24, a typical probe station 20 includes a wafer-receiving table or chuck 28 for supporting the wafer 22. The probe measurement network 21 of the station includes a probing assembly 30 which, as shown, can take the form of a probe card with a multiconductor probe tip array for delivering signals to, and receiving signals from, the respective bonding pads of each individual device. One common type of probe card structure, as depicted, includes an open-centered rectangular-shaped frame 32 with numerous needle-like probe tips 34 that downwardly converge toward the open center of the frame. The end portion of each tip is bent at a predetermined angle so that the lower extremities or device-probing ends of the tips, which typically have been blunted by lapping to form a coplanar array, are suitably arranged for one-to-one contact with the bonding pads 26 provided on each respective microelectronic device. The measurement signals provided by the network are generated within and monitored by a multichannel test instrument 36, which is connected to the probe card via a suitable multiconductor cable 38. The probe station also includes an X-Y-Z positioner (e.g., controlled by three separate micrometer knobs 40a, b, c) for permitting fine adjustments in the relative positions of the probe card 30 and the selected device-under-test.
The individual elements that make up a probe measurement network can take forms other than those shown in FIG. 1. For example, depending on the particular requirements of the devices to be measured, the probing assembly can take the form of a multiconductor coplanar waveguide as shown in Strid et al. U.S. Pat. No. 4,827,211 or Eddison et al. UK Pat. No. 2,197,081. Alternatively, the assembly can take the form of an encapsulated-tip probe card as shown in Higgins et al. U.S. Pat. No. 4,566,184, or a multiplane probe card as shown in Sorna et al. U.S. Pat. No. 5,144,228, or a dual-function probe card in which the probe card not only probes but also supports the downturned wafer, as shown in Kwon et al. U.S. Pat. No. 5,070,297. Use of this last card structure, however, is limited to the probing of flat wafers or other device configurations in which all devices are of the same height.
Before using a probing station or other probing system to measure the high-frequency performance of individual devices, such as those formed on a wafer, it is desirable to first accurately evaluate the signal conditions that are actually present in the measurement network of the system with reference, in particular, to the device-probing ends of the network.
For example, with respect to a probing system of the type shown in FIG. 1, in order to accurately calibrate the source or incoming channels of the system's measurement network, preferably measurements are made of the respective signals that are generated by the various sourcing units of the test instrument 36 in order to reveal how these signals actually appear in relation to each other when they arrive at the device-probing ends that correspond to the respective source channels, since the signals that actually enter the input pads of each device come directly from these ends. Conversely, in order to accurately calibrate the sense or outgoing channels of the probing network, preferably the respective signal conditions that are indicated by the various sensing units of the test instrument 36 are observed when reference signals of identical or otherwise relatively known condition are conveyed to the device-probing ends that correspond to the respective sense channels, since the signals that actually exit the output pads of each device go directly to these ends. Should any channel-to-channel differences be found to exist in the network, these differences can be compensated for so that the test instrument will only respond to those differences which actually arise from the different input/output characteristics of the device-under-test.
Typically it is difficult, however, to make comparatively accurate high-frequency measurements in reference to the extreme ends of a probing assembly where the ends have been arranged for the measurement of planar microelectronic devices because of the reduced size and the closely crowded arrangement of such ends. This is particularly so when the probing assembly is of the card-like type 30 shown in FIG. 1, due to the inherent fragility of the needle-like tips 34 that are part of such an assembly.
The reason for this difficulty can be better understood in reference to FIG. 3, which shows one common type of interconnect assembly that has been used to evaluate probe measurement systems of the type shown in FIG. 1. This assembly includes a signal probe 42 having a single pointed transmission end 44, which probe is connected, via a cable, to the sensing unit, for example, of a test instrument. This instrument can either be the same as that instrument 36 which provides the sourcing units for the probe measurement network or, as shown, can be an entirely separate instrument 46. Viewing FIGS. 1 and 3 together, when the pointed end of the signal probe is being repositioned from one tip 34 to another, normally it is necessary to move the relatively stiff end of the probe slowly and deliberately in order to avoid damaging the delicate needle-like tips, so that a relatively long period of time is needed in order to complete evaluation in relation to all the tips. Additionally, this type of probe has poor high-frequency measurement stability in moderately noisy test environments. Even more significantly, because the extreme ends of the needle-like tips 34 on the probe card are too thin and delicate to be probed directly, probe-to-probe contact between the pointed transmission end 44 of the signal probe and each needle-like tip of the probe card must occur further up nearer to the base of each tip. This introduces, for example, a phase offset of indeterminate amount between the signal that is being measured by the signal probe and the signal as it will actually appear in relation to the bonding pads 26 (FIG. 2) of each device. The degree of this offset, moreover, will generally vary in an arbitrary manner between the different tips, since the pointed end of the probe will normally be placed into contact with the different tips at somewhat different positions along their respective lengths. Using this type of calibration assembly, then, it is difficult, if not impossible, to accurately evaluate the relationships of the different signals that actually exit the various device-probing ends of the needle-like tips 34 and hence it is difficult, if not impossible, to normalize these signals or to otherwise calibrate the network so as to permit accurate device measurement.
An alternative approach to probe network evaluation would be to use one or more of the deviceprobing ends that are included on the probe card itself, instead of a separate signal probe, for establishing the reference channel back to the original test instrument. In accordance with this approach, a different form of interconnect assembly would be used. This assembly might include a plurality of conductive paths, such as those defined by traces formed on a substrate, where the arrangement of the paths would be such that each deviceprobing end relative to which evaluation is to be conducted would be connected to one of the ends being used to establish the reference channel via a "through" channel formed by one or more of the paths.
However, through channels of this type would constitute less than perfect transmission lines and, to the extent that the majority of the ends are to be evaluated in this manner, these through channels would need to be of different lengths to accommodate such measurement. Hence, even when the same source or sense channel is being evaluated under this approach, the measured value of signal condition in the channel will appear to change depending on which through channel of the assembly is being used for making the observation. Moreover, since a typical probe card for wafer-level testing has hundreds of probe ends converging within an area less than one-half inch on each side, and since there can be cross-coupling of signals between closely adjacent paths as well as distortion caused by the presence of extraneous radiation in the measurement environment, a suitable physical layout that could provide, for example, adequate high-frequency signal isolation for each path is not readily apparent.
Although its use is limited to a probe card of quite different type than that shown in FIG. 1, another type of high-speed interconnect assembly which uses a signal probe for evaluating probing networks is described in J. Tompkins, "Evaluating High Speed AC Testers," IBM Technical Disclosure Bulletin, Vol. 13, No. 7, pp 1807-1808 (December 1970). As in Kwon et al., in Tompkins it is the probe card itself that provides support for the device-under-test, that is, the device is turned over so that its bonding pads rest upon a plurality of slightly-raised rounded probing ends included on the upper side of the card. As in Kwon et al., this mounting method forecloses the testing of hybrid devices in which components of different height are mounted on the face of the device. An additional disadvantage of the Tompkin's probing network is the poorly regulated interconductor spacing in the lead-in cable to the card, which can result in signal instability at higher frequencies. In any event, to evaluate the signals that are present in the network in reference to the rounded probing ends on the card, the interconnect assembly of Tompkins includes a two-prong signal probe together with a sheet-like dielectric member which is placed in a predefined position over the device-supporting or upper side of the probe card. Uniformly-spaced holes are formed through the dielectric member and serve as guide channels for guiding the first prong of the signal probe into tip-to-tip contact with the various rounded probing ends on the card. At the same time, a shorter second prong of the signal probe automatically establishes contact with a conductive ground plane which is formed on the upper side of the dielectric member and which surrounds each hole on that member.
There are significant difficulties with the type of evaluation approach just described, however, because the pointed end formed on the first prong of the signal probe can, over time, wear down the rounded ends of the probe card so that these rounded ends eventually lose their capacity to establish simultaneous electrical contact with the planar pads of the device-under-test. Furthermore, this measurement approach does not permit, while device testing is in progress, quick evaluation of signal condition with respect to a particular probing end of the card, because the first prong of the probe normally cannot be applied to any of the ends of the card until after the device has been carefully lifted off the card and removed to a safe static-free location.
Another approach to evaluating the measurement network of a probing system employs an impedance standard substrate of the type described, for example, in Carlton, et al., U.S. Pat. No. 4,994,737. An impedance standard substrate comprises a substrate on which there are known impedance standards, which standards are suitably configured for simultaneous probing by the device-probing ends of the network. The standards can include, for example, an open circuit transmission line element formed by a pair of spaced-apart pads. Unlike the evaluation methods thus far described, no separate reference channel is provided to receive each signal as each signal exits the tip end of a respective incoming channel. Instead, the impedance standard on the substrate is used for reflecting the incoming signal so that the signal is transformed at the tip to an outgoing signal which then travels back to the test instrument through its original signal channel. The electrical characteristics of the corresponding signal channel can then be analyzed from measurements taken at the test instrument using time-domain reflectometry.
However, in a multichannel network, the differences which exist between the incoming signals at the device-probing ends of the various incoming channels are a function not only of the differences which exist in the respective circuit characteristics of those channels (i.e., the differences in the relative conditions for the signals) but are also a function of the differences which exist in the signals themselves from the moment that each is first generated within a respective sourcing unit of the test instrument (i.e., the differences in the respective conditions of the signals). Because the type of evaluation that is made with an impedance standard substrate only detects differences of the former sort and not of the latter, this type of approach, at least by itself, cannot be used to fully evaluate the differences in the incoming signals in reference to the device-probing ends of the measurement network. Conversely, the differences in the signal conditions that are indicated by the various sensing units of the test instrument, even when reference outgoing signals of identical condition are presented to the device-probing ends of the corresponding sense channels, are not observable using the impedance standard substrate approach. Thus, this approach does not permit. the different signal conditions of a multichannel probe measurement network to be fully characterized and compensated for so as to allow accurate device measurement. It may also be noted that expensive processing is normally needed in order to properly evaluate time-domain reflectometry measurements, because the signal which is evaluated in these types of measurements is prone to significant cumulative distortion due to partial reflections occurring along the channel, conductor losses, frequency dispersion and so on.
One type of probe card evaluation system which is unsuitable for high-frequency measurements but which can be used in relation to an array of probe tips for measuring certain low frequency or DC characteristics is sold by Applied Precision, Inc., of Mercer Island, Wash. under the trade name CHECKPOINT.TM.. The design of this system is patented in Stewart et al., U.S. Pat. No. 4,918,374, and a similar system is apparently made by Integrated Technology Corporation of Tempe, Ariz., under the trade name PROBILT PB500A.TM.. As described in Stewart, the evaluating system has its own probe card holder. The probe card is transferred to this holder so that the probe card can be held in a predetermined position above a square-shaped checkplate, the upper side of which is divided into four quadrants. In one characteristic construction, at least one of the quadrants contains a narrow conductive strip extending in either an X or Y reference direction. To determine the X position of a particular tip, for example, the Y directional strip is moved by incremental movements of the underlying checkplate in the X direction toward the tip until a continuity reading between the Y directional strip and the tip reveals the precise X position of that tip relative to the checkplate's original position and hence relative to the card. In order to determine the positions of several tips at the same time, in a second construction, one of the quadrants contains a number of spaced-apart parallel strips that are each wired out to a separate terminal on the sides of the checkplate, thereby making it possible to discern, for purposes of positional verification, which strip is in contact with which tip.
In order to determine the respective positions of two tips that have been electrically tied together at some point up from their ends, yet a third construction is used in Stewart, since under the first two constructions there can apparently be some difficulty in determining visually which particular tip of the two that are tied together is actually in contact with a strip when continuity is detected. In this third construction, one of the quadrants contains a solitary conductive dot of sufficient smallness that only one probe tip at a time can be placed on the dot, thereby enabling the position of each tip to be determined in consecutive sequence. In order to get a proper continuity reading, any other conductor on the checkplate besides the dot is confined to another quadrant of the checkplate. Hence, any other tip that might be tied to the tip under test, including a tip on the opposite side of the card, cannot come into contact with another conductor as the tip under test approaches the dot, which would confusingly produce the same reading as if the tip under test had achieved contact with the dot. For apparently similar reasons, the conductive dot is wired out to a terminal that is separate from the terminal of any conductor in the other quadrants.
From the foregoing description of the Stewart evaluation system, it will be recognized that the principal use of this system is to precisely locate the relative positions of the device-probing ends of the measurement network. Although it might be possible to upgrade the Stewart system to permit the evaluation of certain lower frequency characteristics (such as by adding, perhaps, a lumped capacitor divider network to the Stewart system to measure low-frequency capacitive effects), its structure is wholly inadequate for higher frequency measurements, such as those ranging above 50 MHz.
For example, to the extent that the conductor arrangement in Stewart assumes the form of several parallel strips in closely spaced relationship to each other, if the signal condition in any channel is evaluated via one of these strips, it can appear to vary depending on which strip is used (given that the electrical length between each strip and its corresponding terminal varies from strip-to-strip), on where exactly the device-probing end of the channel is placed in relation to the elongate strip, and on what types of distorting signals are present in the immediate vicinity of the device-probing end (since relatively unrestricted coupling of signals can occur between the closely neighboring strips). Similarly, to the extent that the conductor arrangement in Stewart takes the form of a solitary dot in any one quadrant, if the signal condition in any channel is evaluated via this dot, it may appear to vary due to coupling between tips and due to any movement of equipment in the vicinity of the channel, particularly since this type of conductor arrangement fails to provide adequate constraint of signal ground. That is, the one or more device-probing ends of the network that normally establish a ground return path for the high-frequency signal channels of the network by their connection, for example, with the ground pad or pads of the device under measurement, are afforded no connection sites in the quadrant of the Stewart checkplate containing the solitary dot. For the same reason, the Stewart system is not able to accurately duplicate during the evaluation session the loading conditions that are present during device measurement.
There are additional disadvantages associated with the Stewart procedure insofar as the probe card is removed in Stewart from its original holder and remounted in a separate stand-alone station before evaluation of the probe card begins. Although this remounting procedure allows the Stewart evaluation station to process the signals before they enter the checkplate, such procedure forecloses the possibility of in situ measurement of the network.
Other systems that have been developed for precisely locating the relative position of the device-probing ends of a measurement network are shown in Sigler, U.S. Pat. No. 5,065,092 and in Jenkins et al., U.S. Pat. No. 5,198,756. These systems, like that of Stewart, are inadequate for high-frequency measurement for similar reasons.
In accordance with the foregoing, then, an object of the present invention is to provide an improved system for evaluating the high-frequency characteristics of a probe measurement network with reference, in particular, to the device-probing ends of such network.
A related object of the present invention is to provide an improved interconnect assembly for uniformly transferring high-frequency signals to and from the device-probing ends of a probe measurement network, particularly when such ends are arranged for the measurement of planar microelectronic devices.